PDA

View Full Version : Dishonest Mainboard

DimensionDude
April 3rd, 2006, 08:58 PM
I have two Elpina M919 mainboards, and while searching for info came across the startling revelation that the "write-back cache" doesn't actually do anything. Look at the (poor quality) pictures to see what I mean.

As you can see, the traces connect only between the chips (or, where the chips should be) and don't go anywhere else on the board.

The other board has the chips in place and they each have "Write-Back" printed on them. That board is in a working system so I can't easily get pictures.

And just for the record, the board will take a 486, X5, 5x86 or P24 processor. It has ISA, VLB and PCI slots. The on-board serial ports don't work on either board (a common problem according to what I've read).

It makes me think of a non-functional hood scoop on a car. It looks cool, but it doesn't help performance at all.

Kent
carlsson
April 4th, 2006, 12:38 PM
Maybe if the voltage is high enough, electrons will jump across from one trace to another? ;-)
mbbrutman
April 4th, 2006, 03:31 PM
Did you actually read somewhere that this motherboard is a fraud, or did you actually test it? Motherboards are built in layers and there may be wiring that you are not seeing.
Vlad
April 4th, 2006, 05:27 PM
Unless he cracks the thing in half, we have no real way of knowing if it has internal tracings or not. I believe that it MIGHT be a fraud. Espcially after my little adventure with the ChainTech Company. They didn't manufacture it properly, so it really could be a badly manufactured mobo. It's not like it hasin't happened before.

-V
DimensionDude
April 4th, 2006, 05:46 PM
I read in several different places (mostly forums that I didn't bookmark) that the cache is a fraud. As for the non-functional serial ports, I tried the recommended fix (cut a specific trace in a specific location) on one of the boards but it didn't work. The only way to get a serial port was to use an I/O card.

Also found out that a 100Mbs PCI lan card will cause it to crash in Win95. However, a 10Mbs works fine.

Oh yeah, Elpina manufactured the substrate (blank board), PC Chips was the assembler/marketer.

Kent
Vlad
April 4th, 2006, 05:49 PM
Atleast you have a paper weight, or a conversation peice....

-V
DimensionDude
April 4th, 2006, 06:32 PM
They're not quite paperweights yet (got way too many paperweights). Both boards work. Since it can take 3 different types of cards, I use one of the boards as a test bed.

More info on the board here http://th2chips.freeservers.com/m919/

That's one of the places where I read that the "write-back" cache is bogus. The COAST (Cache On A Stick?) is real, however.
carlsson
April 5th, 2006, 12:49 AM
On the topic of NICs and motherboards, I clearly remember that some types of NICs would not work well in one motherboard, but fabulous in another. And vice versa. In the computer club, we had a diversity of both types, and more than once I juggled two cards between computers to improve the network speed. Sometimes a 10 Mbps ISA card would be more responsive than a 100 Mbps PCI card, until we could find the brand of 100 Mbps NIC that motherboard was compatible with.

We helped students with their computers, and sometimes it happened that a computer would refuse to allocate an IRQ or at least install the NIC of one type. Then we (I) replaced it with another brand but same specs, and it worked perfect. If you have NICs based on several different chipsets, it could be worth trying a different card and hold your breath that it would work better.
Vlad
April 5th, 2006, 03:16 AM
Believe it or not COAST really does mean cache on a stick. I love when they use off the wall terms like that. I looked closer, I see what you mean about the fake chips. Maybe they were wireless? (Traceless?)

-V
Vlad
April 6th, 2006, 02:23 PM
I thought came to me, What if they designed the mobo from a general use template and just didn't use those two chips? Maybe that's what happened?

-V
carlsson
April 6th, 2006, 02:50 PM
You mean several different models of the same basic motherboard, but this particular model didn't implement the write-back cache, even if you installed the chips? How can one identify that nothing happens, checking if the chips get hot?
Vlad
April 6th, 2006, 03:00 PM
You mean several different models of the same basic motherboard, but this particular model didn't implement the write-back cache, even if you installed the chips?

More or less yeah. Kind of a basic board they customized to make their product.

How can one identify that nothing happens, checking if the chips get hot?

I doubt they would do anything since they don't connect to anything but the board. The tracings don't go anywhere.


-V
carlsson
April 6th, 2006, 03:06 PM
But what about the theory with multi layered boards, and traces inside the circuit board?
Vlad
April 6th, 2006, 03:08 PM
Could be I guess. No way to really tell unless he breaks it. But thats a little too far....

Who knows? Maybe its just a strange board.

-V
DimensionDude
April 6th, 2006, 05:02 PM
On the subject of whether or not the cache works, I may have the definitive answer. I put one (the one with the cache chips) in a case and installed Win98SE then ran PassMark PerformanceTest. Then I swapped boards and ran PerformanceTest again. Both boards scored the same.

I suppose that I could find some sort of cache testing software and try that, but I would bet that both would still be the same.

Or, since the board without the chips is not installed in a case, I could spend some time with an ohmmeter and find out for sure if the traces really go nowhere. I'm skeptical about the multi-layer theory as every solder pad has a visible trace that connects to another solder pad.

To expand on carlsson's theory, perhaps the data is transferred to the write back cache by good intentions :) Only data with the necessary karma can make the transition.

Kent
Terry Yager
April 6th, 2006, 05:15 PM
So, if you were to become an evil haxor, and try to steal my CC numbers, then the board would self-destruct?

--T
dpatten
April 7th, 2006, 03:00 PM
PC-chips boards were known for this. One board actually had FAKE ICs. That is they LOOKED like a cache chip, they had metal contacts, but they were basically just a chunk of plastic with dead contacts, manufactured specifically to rip you off.

http://redhill.net.au/b/b-bad.html

Indeed your PC chips 919 board does have fake chips...
scroll down the following page.

http://redhill.net.au/b/b-96.html
DimensionDude
April 7th, 2006, 07:29 PM
dpatten,

Thanks for the links, that seems to confirm what I had seen elsewhere. The M919 shown in the second link is exactly the board I have.

AIDA32 shows no L2 cache, even though the "write-back" chips are in place and the COAST module is installed. At bootup it shows "256K cache enabled."

Funny thing, that site sings the praises of Chaintech. I guess they were good back in the day, but I've seen more than one complaint about new Chaintech boards recently.

Kent
Vlad
April 7th, 2006, 07:31 PM
.......but I've seen more than one complaint about new Chaintech boards recently.

Kent

Tee Hee

Never had an old one.....

-V
carlsson
April 8th, 2006, 05:15 AM
We had a number of different Socket 7 motherboards, many with the cache slot. Once I obtained a cache module, but IIRC it only made the computer more crash prone, definitely not faster or better.
dpatten
April 10th, 2006, 06:05 AM
dpatten,

Thanks for the links, that seems to confirm what I had seen elsewhere. The M919 shown in the second link is exactly the board I have.

AIDA32 shows no L2 cache, even though the "write-back" chips are in place and the COAST module is installed. At bootup it shows "256K cache enabled."

Funny thing, that site sings the praises of Chaintech. I guess they were good back in the day, but I've seen more than one complaint about new Chaintech boards recently.

Kent

Try swapping the BIOS with another board that is substantially similar. Chances are the "256K cache installed" will disappear.